Monday, April 3rd, 2023

import matplotlib.pyplot as plt
import numpy as np

motorbike = plt.imread('motorbike.png')


# or 
#motorbike = plt.imread('motorbike.png')[:,:,0]

motorbike[0,0]

array([0.88235295, 0.88235295, 0.88235295], dtype=float32)

plt.imshow(motorbike)

<matplotlib.image.AxesImage at 0x2604844c3a0>

motorbike.shape

(552, 640, 3)

motorbike[0,0]

array([0.88235295, 0.88235295, 0.88235295], dtype=float32)

Note: the image itself is a grayscale image (meaning the red-values, green-values, and blue-values are equal for each pixel). On the other hand, plt.imread() loaded the image with full RGB values (i.e. there is a third dimension of length 3).

motorbike_gray = motorbike[:,:,0]                # Take just the red-values for our grayscale value

motorbike_gray.shape

(552, 640)

motorbike_gray[0,0]

0.88235295

plt.imshow(motorbike_gray)

<matplotlib.image.AxesImage at 0x260484aab20>

plt.imshow(motorbike_gray, cmap='gray')    # Tell imshow to use a gray colormap

<matplotlib.image.AxesImage at 0x26048688460>

Recall from last week: Use np.random.rand to generate salt and pepper noise:

nrows,ncols = motorbike_gray.shape
pepper_mask = np.random.rand(nrows,ncols) < .05

motorbike_noise = motorbike_gray.copy()
motorbike_noise[pepper_mask] = 0

salt_mask = np.random.rand(nrows,ncols) < .05
motorbike_noise[salt_mask] = 1

plt.figure(figsize=(15,15))
plt.imshow(motorbike_noise, cmap = 'gray')

<matplotlib.image.AxesImage at 0x26049523340>

# This code would give approximately 10% image noise
# 5% from salt noise
# 5% from pepper noise

motorbike_noise = motorbike_gray.copy()

nrows,ncols = motorbike_noise.shape
pepper_mask = np.random.rand(nrows,ncols) < .05
salt_mask = np.random.rand(nrows,ncols) < .05

motorbike_noise[pepper_mask] = 0
motorbike_noise[salt_mask] = 1

plt.figure(figsize=(15,15))
plt.imshow(motorbike_noise,cmap='gray')

<matplotlib.image.AxesImage at 0x2604aeea6a0>

The above code is biased toward salt noise, since we set salt noise second. In other words, we might change previously pepper noised pixels to salt noised pixels.

# This code would give approximately 10% image noise
# 5% from salt noise
# 5% from pepper noise

motorbike_noise = motorbike_gray.copy()

nrows,ncols = motorbike_noise.shape

random_array = np.random.rand(nrows, ncols)

pepper_mask = random_array < .05
salt_mask = random_array > .95

motorbike_noise[pepper_mask] = 0
motorbike_noise[salt_mask] = 1

plt.figure(figsize=(15,15))
plt.imshow(motorbike_noise,cmap='gray')

<matplotlib.image.AxesImage at 0x2604b128970>

def sp_noise(img, noise):
    ...
    return img_noise

filtered_img = np.zeros_like(img)

for i in range(1,nrows-1):
    for j in range(1,ncols-1):
        ...
        filtered_img[i,j] = 

monkey = plt.imread('monkey.jpg')

plt.imshow(monkey)

<matplotlib.image.AxesImage at 0x2604cad3cd0>

monkey_gray = np.mean(monkey,axis=2)

plt.imshow(monkey_gray,cmap = 'gray')

<matplotlib.image.AxesImage at 0x2604c9baf10>